Applied Scientist 5

Adobe Adobe · Enterprise · Bangalore, India +1

Senior Machine Learning Engineer with expertise in generative modeling (diffusion models) and computer vision to join Adobe's Applied AI team. The role involves architecting and shipping state-of-the-art diffusion-based models, driving applied research into production, and mentoring engineers. Responsibilities include designing, training, and fine-tuning diffusion models for image/video generation, improving sampling efficiency, building production pipelines for image/video understanding, owning the ML lifecycle, optimizing models for inference, and designing scalable training infrastructure. The role also involves defining evaluation frameworks and leading technical design reviews.

What you'd actually do

  1. Design, train, and fine-tune large-scale diffusion models (DDPM, DDIM, LDM, DiT) for image, video, and multimodal generation tasks.
  2. Build production-grade pipelines for image/video understanding: segmentation, detection, depth estimation, optical flow, and 3D reconstruction.
  3. Own the full ML lifecycle: data curation, experiment tracking, model evaluation, optimization, deployment, and monitoring.
  4. Optimize models for inference: quantization (INT8/FP8), ONNX export, Flash Attention, and xFormers.
  5. Design scalable training infrastructure on distributed GPU clusters (DDP, FSDP, DeepSpeed) across thousands of GPU-hours.

Skills

Required

  • Python
  • PyTorch
  • score-based and diffusion models
  • computer vision fundamentals
  • fine-tuning large vision and generative models
  • distributed training frameworks (DDP, FSDP, DeepSpeed, Megatron-LM)
  • probabilistic ML, variational inference, and information theory
  • MLOps tooling (Weights & Biases, MLflow, DVC, or equivalent)
  • shipping ML models to production

Nice to have

  • flow-based generative models
  • video generation models
  • 3D generative models
  • multimodal systems
  • RLHF / DPO for generative model alignment
  • Active open-source contributions
  • Active GitHub presence

What the JD emphasized

  • deep expertise in generative modeling and computer vision
  • architect and ship state-of-the-art diffusion-based models
  • drive applied research into production
  • translating the latest advances in generative AI into scalable, reliable systems
  • Deep theoretical and practical knowledge of score-based and diffusion models
  • Experience fine-tuning large vision and generative models at scale
  • Track record of shipping ML models to production at scale

Other signals

  • design, train, and fine-tune large-scale diffusion models
  • build production-grade pipelines for image/video understanding
  • own the full ML lifecycle
  • optimize models for inference
  • design scalable training infrastructure